Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Opukai meant to stop evaluating AI models before they were released for the danger they could attract propaganda campaigns or making more effective propaganda campaigns.
The company says today will discuss the risks through the terms of the service, which prevents the use of AI models in political campaigns and monitor how people are used in signs of violations.
Opuki also said to think about releasing AI models that are judged “high risk” when a rival AI label has already released the same model. Previously, Opasii said it would not be released to any AI model presented more than a “medium risk.”
Policy changes have been placed in an update to Openia’s “Prepare the framework ” Yesterday. Those details details are how companies monitor AI models have established for a person’s risk of making hackers in the possibility that models will develop self and escape human control.
Policy changes dividing safety and security experts. Many carry on social media to praise OpenI for voluntary release of updated framework, stronger risk categories such as threats of threats such as autonomous evacuation and protection prevention.
However, some express concerns, including Steven Adler, a former Safety Safety of Openia who criticized the fact that the updated tests no longer need survival tests. “Openi is quiet to reduce its safety reasons,” he wrote X. However, he emphasized that he appreciated Openia’s efforts: “I am generally happy to see preparation to prepare updated,” he said. “It’s likely to be a lot of work, and never needed.”
Some critics promote removal of persuasion from disasters owned by preparations of versions.
“Openi appears to be moved in this way,” Said Shyam Krishna, a research leader in AI and Rand Europe management. “Instead of being treated attraction as a risk category, it can now be a higher level of socy-societal and integrated guidelines for opening advances.” It remains to see how it plays in politics, he added, where dazzling AI “is still a controlled issue.”
Courtney Radsch, a senior partner of the streams, the center for international administration changes, and the center for democracy and technology that still works in a message in wealth “Another instance of technology hubricity technology.” He emphasized that the decision of the ‘persuasion’ “does not pay attention to the context – for example, attracting may not be unreasonable to children or the AI states.”
Oren Etzioni, former CEO of Allen Institute for AI and Founder of Treemedia, offering tools to fight a manipulated content, the anxiety is also declared. “Depreciation to deceive me as an error given to the growing power of LLMS,” he said to an email. “One should think if OpenI simply focuses on the income with little esteem for the impact of society.”
However Ai researcher at AI does not associate with Opui’s wealth It seems reasonable that responding to any risks from disinformation or other harmful persuasion used through OPUII’s service terms. The researcher, asking to remain anonymous because he is not allowed to speak public without permission from his current employer advivation in pre-deployment testing. Besides, he indicated that this category is dangerous is more amorphus and ambivalent compared to other critical or biological attacks of firearms or biological a person in a cyberattack.
Noted that some members of the European Parliament have Also expressed concern That the most recent draft of the proposed practice code for following EU AI Acgidred mandatory test of AI models for the possibility that they can spread the disinferation of a voluntary consideration.
Studies find AI Chatbots to be more convincing, even if this self-worth does not need to be dangerous. Cornell University researchers and MIT, for example, found Such dialogues with chatbots are effective in getting people in question with conspiracy theories.
Another criticism of the updated OpenII framework is centered on a line where OpenII says: “If a Frontier AI developer releases our needs.”
Max Tegmark, the President of the Future of Life Institute, a non-profit attempt to solve those who have AI’s hazards, a statement of wealth That “The Race to the bottom is speeding up. These companies are openly racing to build uncontrollable artificial General Intelligence-Smarter-Than-Human AI Systems Designed to Replace Humans-despite admitting the massive risks this poses to our workers, our families, our national security, Even our continued existence. “
“They say no one of what they say about the safety of AI engraved on the stone,” as the longest critique opui LinkedIn Message, who said the line was mainly in a race at the bottom. “The ruler of their decisions is the competition pressure – not safe. Slowly they promise a securing private data – instead of a non-prisoned Social Data – instead of a nonprofit-directed data-spoken of private data – Instead of a nonprofit pointed data-surves private data – instead of a non-competing Social Survey of a nonprofit pointing private data – than their suggested social prompting.
Overall, it is useful that companies like Openi share their thinking around their risk management behaviors, Director of Democracy and Technology, told wealth in an email.
That is said, he added that he was worried about moving goutposts. “It is a disturbing trend if, as AI systems seem to be bad risks, those risks themselves are in the contents of guides,” he said.
He also criticizes the focus of the ‘Frontier’ models when used by OpenII and other technical companies not to publish safety surveys of new, strong models. (Eg, openiai released The 4.1 model of this unsafe report, says this is not a covered model). In other cases, companies have any Failed to publish safety reports Or slow to do so, publish them in months after releasing the model.
“Among the kinds of issues and an advancing pattern of AI developers where new models have been launched by their own releases,” he said voluntary commitments
Update, April 16: This story is updated to include a comment from the future of life Institute President Max Tegmark.
This story originally shown Fortune.com
